Another change the audio engineer faces when working with 3D sound is how to deal with post processing of 3D processed audio.
Ideally the 3D audio process is the last process in the effects chain. The binaural output of the 3D process should not undergo any further audio processing.
This means that all your other effects like EQ, time based effects (delay, reverb, modulation), dynamics (compression, limiting) should happen before the 3D input. This also means that all these pre-processes are done in mono. (see blog #2)
However, in a complex game or film mix you might work with 100’s of sounds/tracks, and these are often at some stage in the mix grouped together in stems (ambience, dialogue, sfx), so that they can be controlled with one fader or level command. Ambience tracks are a common example. These are often side chained so that their level depends on what other sounds are played in the scene.
So although in an ideal world you wouldn’t apply an post processing to 3D sounds, sometimes reality dictates differently. Below an overview of ‘damage control’ and do’s and don’ts.
Because the 3D processing generates two new channels (left and right) based on the mono input, and the differences in the output between left and right are both in phase/timing and frequency levels. These differences between left and right are specific for the location where you placed the sound source in relation to the listener. For every movement both the phase/timing and frequency of the mono input is changed for left and right to mimic the new position in 3D.
As long as you keep that in mind you’ll realise that any change you make after the processing, might mess up the 3D experience.
But some processes are worse than others:
Whenever you can, you should apply EQ before you go through the 3D process. However, especially the very low frequencies are not super important for localisation, so if needed you can use shelving eq’s, but stay below 100Hz. Also use equalisers that behave as linear as possible, so no emulation of analogue hardware or filters with strong resonance peaks, but clean EQs with gentle slopes. Be careful with the high frequencies (>5kHz) as these still help in localising the source, also because higher frequencies are more directional and will vary more when changing position.
Effects like chorus, flanging or modulating delays should be avoided at any cost after the 3D processing. Especially if the L and R side are modulated independent you will definitely harm or even ruin the localisation. Reverbs can be used with care, if possible apply reverb before you apply 3D processing. However, especially if the reverb contains some pre delay before the tail kicks in, you can add some reverb after the 3D processing if needed. Make sure you keep the level relatively low, so that it doesn’t obscure the positional cues our 3D engine generates.
Same as time based effects. They both change the frequency content as well as the timing of the signal, and should not be used after 3D processing
Although ideally you should apply compression before the 3D processing, there is often the need for some bus compression. And in those cases the bus might be side chained. (for instance to automatically lower the level of a helicopter sound or a group of ambience tracks when there is dialogue)
The good news is that if used with care, 3D processing will survive compression. The trick is in the use of the attack and release times:
The longer the attack and release time, the less harm (if any) will be done to the 3D effect.
This is because with longer times a compressor is more an automated volume control than a ‘wave shaper’.
What is short, what is long?
If you look at a 40Hz waveform then you can easily calculate that one cycle takes 1000/40= 25 ms.
if your sounds go down to 20Hz (but that’s very low, most headphones won’t play them anyway) then one cycle would take 50ms.
If your attack/release times are below these numbers, you’re reshaping the wave form. As long as you stay above (the longer, the better) your compressor will function more like a volume control, and the 3D effect won’t suffer from changes in volume, as long as they are gradual.
Bear in mind that the volume changes should be kept relatively small, but that is common when used in side chaining. This is because volume is also associated with distance, and too obvious volume changes will be interpreted by the viewer/player as a change in distance. (6dB level drop is a doubling of distance)
Having at least a limiter on your master output bus is common in digital audio; you want to avoid clipping your DA converters.
Limiting is a different story than compression; digital limiters often have extreem short attack and release time and immediately lower the volume for even a one sample peak above the threshold. This will change the waveform, and might therefor reduce the 3D effect.
However, this would only be the case during gain reduction, so if you inserted a limiter only to reduce the loudest peaks, then only during those peaks the 3D effect will suffer, which might in reality be unnoticeable. The more you push your limiter, and the more it is working, the more the 3D effect will suffer.
In other words, gentle limiting used as protection for overload will work fine, while ‘pushing’ a limiter to create an ‘in your face’ sound will harm the 3D effect.
We hope the above helps a bit in sculpting your 3D mix. If you have any questions or remarks, feel free to contact me.